triflux@10.41.1
CLI-first multi-model orchestrator for Claude Code — route tasks to Codex, Antigravity, and Claude
AI Security Review
scanned 4mo ago · by lpm-firewall-aiStatic inspection found powerful CLI behavior, but it is consistent with triflux's advertised AI-orchestration purpose and is activated by explicit user commands rather than npm lifecycle or import-time code. No credential theft, destructive behavior, persistence abuse, or attacker-controlled exfiltration endpoint was confirmed.
Decision evidence
public snapshot- bin/triflux.mjs and scripts/setup.mjs can modify Claude/Codex/MCP config when user runs setup/doctor --fix.
- scripts/setup.mjs can register Claude hooks and statusLine commands for triflux orchestration.
- bin/triflux.mjs has user-invoked update/doctor paths that run npm, git, brew, or local CLIs.
- package.json defines bin entrypoints but no preinstall/install/postinstall lifecycle scripts.
- Network use is package-aligned: loopback hub health/MCP, GitHub star/update checks, npm registry update checks, and configured MCP endpoints.
- hub/team/synapse-http.mjs posts only to configured/loopback hub and uses TFX_HUB_TOKEN or ~/.claude/.tfx-hub-token for local hub auth.
- No source evidence of credential harvesting or exfiltration to attacker-controlled endpoints.
- Shell/child_process use is CLI-driven for diagnostics, setup, tmux/ssh orchestration, update, and optional dependency repair.
- AI-agent config writes are tied to explicit setup/doctor/sync commands, not install-time or import-time execution.
Source & flagged code
9 flagged · loading sourceA single source file combines environment access, network access, and code or shell execution; review context before blocking.
Findings
Affected versions and remediation
This report applies to triflux@10.41.1.
See all versions of triflux for other recorded verdicts.
- Review the evidence and your use of triflux@10.41.1 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.